SAR variant target identification method based on local textural feature

Abstract

The invention discloses an SAR variant target identification method based on local textural features, mainly solving the problem of low SAR variant target identification rate with the existing identification method. The realization process is as follows: 1. improving the statistical distribution of each part of the SAR target by partial differential; 2. cutting a target part for the SAR target which is performed with partial differential transform by Otsu; 3. rotating the target to 90 degrees, selecting a sliding window with fixed size, and cutting in different directions according to different azimuth angles; 4. carrying out Gabor transform to the cut SAR target; 5. coding each image performed with Gabor transform by an LBP operator, and building a histogram; 6. matching a test sample with each SAR image of the training sample by the histogram, abandoning parts with small matching result, and only reserving parts with good matching result; and 7. judging an identification result witha nearest neighbour method. The invention can utilize the local textural features to improve the identification rate of the SAR target variant and is used for identifying a terrain object.

Description

technical field [0001] The invention belongs to the technical field of radar, in particular to a SAR target recognition method, which can be used for ground stationary target recognition. Background technique [0002] In the field of SAR target recognition, in order to judge the objectivity of the recognition algorithm, the experimental data is selected from the actual SAR ground stationary military target data provided by the US DARPA / APERL MSTAR project team. The test data is divided into two categories. The training samples are the imaging data of SAR on the ground when the pitch angle is 17°, including three types of targets: T72sn_132, BMP2sn_c21, and BTR70sn_c71. In order to verify the generalization and practicability of the algorithm, the test sample is the imaging data of SAR on the ground when the pitch angle is 15°, including 7 models in 3 categories, of which two models are added to T72: T72sn_812, T72sn_S7, and BMP is added Two models: BMP2sn_9563 and BMP2sn_95...

AI Technical Summary

Problems solved by technology

This method is to compare the test sample with the standard template formed by the training sample according to a certain matching criterion, so as to complete the recognition of the test target, but the recognition rate is relatively low.

In 2004, Han Ping et al. proposed a template matching method based on segmentation in the paper SAR Image Target Feature Extraction and Recognition Based on KPCA Criteria. This method first preprocesses the SAR image, accumulates all images in a certain orientation unit and obtains the mean value as Template, using the distance measurement method to compare the similarity of the two images, the recognition rate is 94.50%, because it needs to accumulate images as a template, it is sensitive to local changes in test data in practical applications

In 2006, Sun Y J proposed the Adaptive Boosting method in the article Adaptive Boosting for SAR Automatic Target Recognition. This method first uses a sliding window to segment the SAR image, then extracts the unsegmented and segmented two-dimensional Fourier transforms, and finally uses AdaBoost for recognition. , this method has a very high recognition rate for the case where the training data and the test data are consistent, up to 100%, but if there are variants in the test data, the recognition rate will drop to 96.12%. , if the test data changes, the AdaBoost parameters need to be retrained

In 2008, Huan Ruohong worked on a new method of synthetic aperture radar image feature extraction and target recognition. The paper proposed based on PCA to extract main features and then use SVM for recognition. The recognition rate is 96.92%. There is also the need to retrain every time the data changes. data problem

In 2009, in the article SAR target feature extraction and recognition based on two-level 2DPCA, Hu Liping proposed to segment first, then use PCA to extract the principal components, and finally use the nearest neighbor method for matching. The recognition rate is 96.41%, and the recognition rate of this method is relatively high. , but since the principal component extracted by PCA is based on the whole, for variants, the recognition rate based on the whole will also decrease due to the local change.

However, in the model-based method, it is necessary to model the SAR image or the SAR image feature vector, and the construction of the model requires a high level of theory and calculation. At present, the recognition rate based on the model is very low. The method of the model is not very practical in practice

[0006] The above-mentioned template matching method and model-based method are both based on the recognition of the overall target. Although the training data and the test data are consistent, although better recognition results can be obtained, due to the needs of actual warfare, many targets are modified. , bunker, barrel rotation, etc., the local part of the test data has changed, which is different from the data trained by the initial training data, that is, there are variants in the test data that are different from the training data. In this case, The recognition rate of the above methods will be significantly reduced

Images